China’s AI Surge: Meituan Open-Sources 1.6T Model
China’s artificial intelligence sector is experiencing a remarkable acceleration across multiple fronts, with major developments spanning open-source AI models, robotics manufacturing, and global governance. The past week alone has seen the open-sourcing of a frontier-level 1.6 trillion-parameter AI model trained entirely on domestic chips, a government announcement that humanoid robot production is on track to exceed 100,000 units this year, and preparations for the 2026 World Artificial Intelligence Conference (WAIC) in Shanghai.
Meituan’s LongCat-2.0: A Milestone for Domestic Chip AI
On July 6, Chinese tech giant Meituan open-sourced LongCat-2.0, a 1.6 trillion-parameter Mixture-of-Experts (MoE) AI model trained entirely on 50,000 domestic Chinese AI ASIC processors — with no Nvidia GPUs used. The model, released under the permissive MIT License for free commercial use, represents a landmark achievement in China’s efforts to build frontier AI capabilities under the constraints of US export controls.
LongCat-2.0 features a 1-million-token context window and is designed for agentic coding — code understanding, generation, and execution. According to benchmark data, it outperforms OpenAI’s GPT-5.5 and Anthropic’s Claude Opus 4.6 in programming and mathematical reasoning, though it lags behind Claude Opus 4.8 in complex multi-disciplinary research. The model’s architecture activates an average of 48 billion parameters per token, with dynamic activation ranging from 33 billion to 56 billion depending on query complexity.
An early preview version, codenamed “Owl Alpha,” was released anonymously in April 2026 and became one of the most widely used free systems on OpenRouter’s API routing platform, accounting for approximately 10.1 trillion monthly tokens. As VentureBeat reported, this successful deployment of alternative silicon “signals a profound structural shift” that could threaten Nvidia’s dominance in AI training hardware.
Huawei Technologies Co., Moore Threads, and MetaX have all confirmed their hardware supports LongCat-2.0, signaling a growing ecosystem of domestic chip alternatives.
Corporate Context
Meituan’s AI push has been years in the making. The company acquired AI startup Light Year for $285 million in 2023, and its R&D spending rose 22% year-on-year to 7 billion yuan (~$1 billion) in Q1 2026. Shares closed up 4.7% at HK$74.95 on the announcement, though they remain approximately 85% below their 2021 peak.
Humanoid Robot Production: 100,000 Units and Counting
At a press conference on July 7, Gan Xiaobin, Deputy Director of MIIT’s Department of Science and Technology, announced that China’s humanoid robot production is expected to exceed 100,000 units this year. “China’s large models and intelligent agents are accelerating iteration,” Gan said, describing AI as a “key variable” becoming a “strong increment” for high-quality economic development.
The announcement came alongside details of the “2026 Special Action for Humanoid Robot and Embodied Intelligence Real-Scene Training,” a joint initiative by MIIT and SASAC launched in June 2026, targeting thousand-unit deployments across 100+ high-value application scenarios. Currently, over 30% of China’s large industrial enterprises now use AI applications.
Shanghai-based startup Zhiyuan Innovation (智元创新), founded in 2023, has already set a global industry record by producing its 15,000th embodied intelligent robot on June 28. Its Jingling G2 model has been deployed for automated sorting, packing, clothing folding, and even calligraphy. Yao Maoqing, Partner and SVP at Zhiyuan, noted that humanoid robots are expected to enter factories at scale within two years, emphasizing that robots need “not only a good body but also a brain — good large model algorithms to understand the environment and tasks.”
WAIC 2026: Shanghai Prepares for Global AI Gathering
The 2026 World Artificial Intelligence Conference (WAIC) is scheduled for July 17-20 in Shanghai, as CCTV News reported. Now in its ninth edition, the conference will feature over 1,100 enterprises, 3,000+ exhibits, and more than 300 global premieres. The exhibition area has surpassed 100,000 square meters for the first time, with dedicated tracks for smart computing and embodied intelligence each gathering over 200 enterprises.
The conference, organized by China’s Ministry of Foreign Affairs, NDRC, MIIT, and seven other major government bodies, will host 140+ themed forums and expects over 1,400 international guests. According to the Shanghai government, the theme is “Intelligent Partners, Creating the Future Together” (智能伙伴 共创未来).
China’s AI industry has reached a core scale of over 1.2 trillion yuan as of 2025, with nearly 200 key standards developed and the AtomGit open-source community now hosting over 11 million registered developers. China ranks first globally in model downloads.
China’s UN Push for Global AI Governance
On the diplomatic front, Shen Jian, Deputy Representative of China’s Permanent Mission to the UN Office in Geneva, called for AI to become a true driver of shared global development during the launch of the Global AI Capability Development Network on July 5. China advocates advancing AI governance under the UN framework, ensuring equal participation and shared benefits for all countries, with a particular focus on bridging the “intelligence divide” for developing nations.
Analysis: What This Means
The convergence of these developments signals a strategic inflection point for China’s AI ambitions. Meituan’s LongCat-2.0 demonstrates that Chinese companies can achieve near-frontier AI performance using domestic chips, directly challenging the effectiveness of US export controls on advanced semiconductors. The success of domestic chip training may accelerate the decoupling of Chinese AI from Western semiconductor supply chains.
Meanwhile, China’s ability to scale humanoid robot production to 100,000 units signals a manufacturing advantage that could reshape global robotics markets. The MIIT’s targeted programs bridging AI research with industrial application — from factory production lines to retail automation — suggest AI is moving from experimental phases into operational reality across China’s manufacturing sector.
On governance, China’s push for multilateral AI oversight under the UN framework positions Beijing as a leader for the Global South, potentially competing with Western-led governance initiatives.
What to Watch For
As WAIC 2026 approaches, several questions remain: Will other Chinese tech giants follow Meituan’s open-source approach? How will the US respond to evidence that export controls are being effectively circumvented? And can China’s humanoid robot production targets be met in practice? The answers will shape not just China’s AI trajectory, but the global balance of technological power.